Multi-Hop QA on HotPotQA
65.6Exact MatchCoopRAG (GPT-4o-mini)
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| CoopRAG (GPT-4o-mini)Generator Backbone=GPT-4o-mini2025.12 | 65.6 | 78.9 | — | — | — | |
| CoopRAG (Gemma2-27B)Generator Backbone=GPT-4o-mini2025.12 | 64.9 | 79.5 | — | — | — | |
| CoopRAG (Llama3.3-70B)Generator Backbone=GPT-4o-mini2025.12 | 64.7 | 79 | — | — | — | |
| CoopRAG (Gemma2-9B)Generator Backbone=GPT-4o-mini2025.12 | 64.4 | 78.1 | — | — | — | |
| NV-Embed-v2 (7B)Generator Backbone=Llama3.3-70B2025.12 | 62.8 | 75.3 | — | — | — | |
| HippoRAG2Generator Backbone=Llama3.3-70B2025.12 | 62.7 | 75.5 | — | — | — | |
| HopRAGGenerator Backbone=GPT-4o-mini2025.12 | 62 | 76.1 | — | — | — | |
| SiReRAGGenerator Backbone=GPT-4o-mini2025.12 | 61.7 | 76.5 | — | — | — | |
| GritLM-7BGenerator Backbone=Llama3.3-70B2025.12 | 60.7 | 73.3 | — | — | — | |
| HiLightSolver=Qwen3-14B, Actor model=Qwen3-4B, Highlight budget (γ)=0.152026.04 | 60.6 | 74.1 | — | — | — | |
| HELPBackbone=Qwen3-30B-A3B-Instruct-25072026.02 | 60.4 | 73.9 | — | — | — | |
| ToPG-Localmax-iter=32026.01 | 59.3 | 72.7 | — | — | — | |
| BFRSSolver=Qwen3-14B2026.04 | 59.3 | 70 | — | — | — | |
| SDGA-PhaseFramework=ZeroSearch, Model=Qwen2.5-7B, Search Depth (Smax)=52026.05 | 59.2 | — | — | — | — | |
| OPROSolver=Qwen3-14B2026.04 | 59.1 | 72.3 | — | — | — | |
| GTE-Qwen2-7B-InstructGenerator Backbone=Llama3.3-70B2025.12 | 58.6 | 71 | — | — | — | |
| HippoRAG2Backbone=Qwen3-30B-A3B-Instruct-25072026.02 | 58.4 | 72.9 | — | — | — | |
| APESolver=Qwen3-14B2026.04 | 58.1 | 72.2 | — | — | — | |
| PRLSolver=Qwen3-14B2026.04 | 57.1 | 71.9 | — | — | — | |
| RAPTORGenerator Backbone=Llama3.3-70B2025.12 | 56.8 | 69.5 | — | — | — | |
| DSPy (MIPROv2)Solver=Qwen3-14B2026.04 | 56.5 | 72.1 | — | — | — | |
| HippoRAG2Generator Backbone=GPT-4o-mini2025.12 | 56.3 | 71.1 | — | — | — | |
| SDGA-AutoFramework=ZeroSearch, Model=Qwen2.5-7B, Search Depth (Smax)=52026.05 | 56.2 | — | — | — | — | |
| MISolver=Qwen3-14B2026.04 | 55.6 | 70.6 | — | — | — | |
| ToPG-Localmax-iter=12026.01 | 55.3 | 67.8 | — | — | — | |
| GraphRAGGenerator Backbone=Llama3.3-70B2025.12 | 55.2 | 68.6 | — | — | — | |
| HippoRAG 22026.01 | 55.1 | 66.9 | — | — | — | |
| HyperGraphRAGBackbone=Qwen3-30B-A3B-Instruct-25072026.02 | 54.3 | 70.2 | — | — | — | |
| SDGA-PhaseFramework=ZeroSearch, Model=Qwen2.5-3B, Search Depth (Smax)=52026.05 | 52.8 | — | — | — | — | |
| Vanilla-RAG2026.01 | 52.7 | 65.5 | — | — | — | |
| HippoRAGGenerator Backbone=Llama3.3-70B2025.12 | 52.6 | 63.5 | — | — | — | |
| BM25Generator Backbone=Llama3.3-70B2025.12 | 52 | 63.4 | — | — | — | |
| SDGA-AutoFramework=ZeroSearch, Model=Qwen2.5-3B, Search Depth (Smax)=52026.05 | 51.5 | — | — | — | — | |
| GraphRAGGenerator Backbone=GPT-4o-mini2025.12 | 51.4 | 67.6 | — | — | — | |
| ContrieverGenerator Backbone=Llama3.3-70B2025.12 | 51.3 | 62.3 | — | — | — | |
| GTR (T5-base)Generator Backbone=Llama3.3-70B2025.12 | 50.6 | 62.8 | — | — | — | |
| RAPTORGenerator Backbone=GPT-4o-mini2025.12 | 50.6 | 64.7 | — | — | — | |
| GRPO-FullFramework=ZeroSearch, Model=Qwen2.5-7B2026.05 | 50.6 | — | — | — | — | |
| ToPG-Naive2026.01 | 49.2 | 61 | — | — | — | |
| LightRAGmode=local2026.01 | 48 | 59.9 | — | — | — | |
| GraphRAGmode=local2026.01 | 47.3 | 60.2 | — | — | — | |
| SDGA-PhaseFramework=Search-R1, Model=Qwen2.5-7B, Search Depth (Smax)=52026.05 | 47.2 | — | — | — | — | |
| HippoRAGGenerator Backbone=GPT-4o-mini2025.12 | 46.3 | 60 | — | — | — | |
| GraphRAG-RouterSystem=Our Method2026.03 | 46.1 | — | — | — | — | |
| GRPO-FullFramework=ZeroSearch, Model=Qwen2.5-3B2026.05 | 44.5 | — | — | — | — | |
| MemPO2026.02 | 42.9 | 57.64 | 0.34 | 15 | — | |
| GRPO-HalfFramework=ZeroSearch, Model=Qwen2.5-7B2026.05 | 42.4 | — | — | — | — | |
| GRPOmemory=none2026.02 | 42.3 | 57.29 | 0.62 | 27 | — | |
| SDGA-PhaseFramework=Search-R1, Model=Qwen2.5-3B, Search Depth (Smax)=52026.05 | 41.2 | — | — | — | — | |
| SDGA-AutoFramework=Search-R1, Model=Qwen2.5-7B, Search Depth (Smax)=52026.05 | 40.2 | — | — | — | — | |
| LinearRAGBackbone=Qwen3-30B-A3B-Instruct-25072026.02 | 39.6 | 52.4 | — | — | — | |
| HippoRAG2System=RAG-Based2026.03 | 39.5 | — | — | — | — | |
| ReSeekBackbone=Qwen2.5-7b-Instruct2025.10 | 38.9 | — | — | — | — | |
| SDGA-AutoFramework=Search-R1, Model=LLaMA-3.2-3B-Base, Search Depth (Smax)=52026.05 | 38.8 | — | — | — | — | |
| TaSR-RAGBackbone=Qwen2.5-72B-Instruct2026.03 | 38.7 | 48.5 | — | — | — | |
| GRPO-FullFramework=Search-R1, Model=Qwen2.5-7B2026.05 | 38.6 | — | — | — | — | |
| LinearRAGSystem=RAG-Based2026.03 | 38.5 | — | — | — | — | |
| DeepResearcher2026.02 | 38.4 | 51.96 | 1.01 | 29 | — | |
| SDGA-PhaseFramework=Search-R1, Model=LLaMA-3.2-3B-Base, Search Depth (Smax)=52026.05 | 38.3 | — | — | — | — | |
| SDGA-AutoFramework=ZeroSearch, Model=LLaMA-3.2-3B-Base, Search Depth (Smax)=52026.05 | 38.2 | — | — | — | — | |
| SDGA-PhaseFramework=ZeroSearch, Model=LLaMA-3.2-3B-Base, Search Depth (Smax)=52026.05 | 37.9 | — | — | — | — | |
| HippoRAGBackbone=Qwen2.5-72B-Instruct2026.03 | 37.2 | 48.5 | — | — | — | |
| Search-R1Backbone=Qwen2.5-7b-Instruct2025.10 | 37 | — | — | — | — | |
| GRPO-HalfFramework=ZeroSearch, Model=Qwen2.5-3B2026.05 | 36.9 | — | — | — | — | |
| Router-R1System=RL-based Agent2026.03 | 36.8 | — | — | — | — | |
| GRPO-FullFramework=Search-R1, Model=Qwen2.5-3B2026.05 | 35.9 | — | — | — | — | |
| TaSR-RAGBackbone=Qwen2.5-7B-Instruct2026.03 | 35.1 | 42 | — | — | — | |
| SDGA-AutoFramework=Search-R1, Model=Qwen2.5-3B, Search Depth (Smax)=52026.05 | 34.8 | — | — | — | — | |
| ZeroSearchBackbone=Qwen2.5-7b-Instruct2025.10 | 34.6 | — | — | — | — | |
| qwen2.5mode=ReAct2026.02 | 34.3 | 47.12 | 1.07 | 26 | — | |
| ReSearch2026.02 | 33.9 | 50.26 | 0.4 | 23 | — | |
| GRPO-FullFramework=ZeroSearch, Model=LLaMA-3.2-3B-Base2026.05 | 33.2 | — | — | — | — | |
| RAGBackbone=Qwen2.5-72B-Instruct2026.03 | 32.8 | 46.2 | — | — | — | |
| Search-o1System=Training-free Agent2026.03 | 32.8 | — | — | — | — | |
| ReSeekBackbone=Qwen2.5-3b-Instruct2025.10 | 32.8 | — | — | — | — | |
| Vanilla RAGSystem=RAG-Based2026.03 | 32.6 | — | — | — | — | |
| Search-R1Backbone=Qwen2.5-3b-Instruct2025.10 | 32.4 | — | — | — | — | |
| GRPO-FullFramework=Search-R1, Model=LLaMA-3.2-3B-Base2026.05 | 31.8 | — | — | — | — | |
| GRPO-HalfFramework=Search-R1, Model=Qwen2.5-7B2026.05 | 31.8 | — | — | — | — | |
| GraphSearchSystem=Training-free Agent2026.03 | 31.7 | — | — | — | — | |
| GRPO-HalfFramework=Search-R1, Model=Qwen2.5-3B2026.05 | 31.5 | — | — | — | — | |
| CoTSystem=Basic LLM2026.03 | 31.3 | — | — | — | — | |
| GraphRAGBackbone=Qwen2.5-72B-Instruct2026.03 | 31.2 | 42.7 | — | — | — | |
| RAPTORSystem=RAG-Based2026.03 | 30.7 | — | — | — | — | |
| RAGBackbone=Qwen2.5-7b-Instruct2025.10 | 29.9 | — | — | — | — | |
| Graph-R1System=RL-based Agent2026.03 | 29.8 | — | — | — | — | |
| HyperGraphRAGBackbone=Qwen2.5-72B-Instruct2026.03 | 29.6 | 42.8 | — | — | — | |
| GraphRAGSystem=RAG-Based2026.03 | 29.2 | — | — | — | — | |
| GRPO-HalfFramework=ZeroSearch, Model=LLaMA-3.2-3B-Base2026.05 | 28.9 | — | — | — | — | |
| GRPO-HalfFramework=Search-R1, Model=LLaMA-3.2-3B-Base2026.05 | 27.8 | — | — | — | — | |
| StructRAGBackbone=Qwen2.5-72B-Instruct2026.03 | 27.5 | 41.8 | — | — | — | |
| ZeroSearchBackbone=Qwen2.5-3b-Instruct2025.10 | 27.4 | — | — | — | — | |
| Direct InferSystem=Basic LLM2026.03 | 26.9 | — | — | — | — | |
| CoTBackbone=Qwen2.5-72B-Instruct2026.03 | 26.4 | 37.2 | — | — | — | |
| ARAMBackbone=Dream-v0-Instruct-7B2026.03 | 25.8 | 36.1 | — | — | — | |
| RAGBackbone=Qwen2.5-3b-Instruct2025.10 | 25.5 | — | — | — | — | |
| HyperGraphRAGSystem=RAG-Based2026.03 | 25.3 | — | — | — | — | |
| Dream + RAGBackbone=Dream-v0-Instruct-7B2026.03 | 25 | 35.2 | — | — | — | |
| Dream + SPREADBackbone=Dream-v0-Instruct-7B2026.03 | 25 | 34.7 | — | — | — | |
| Search-R1System=RL-based Agent2026.03 | 24.7 | — | — | — | — |